混响环境下改进的球谐波域L1-SVD声源定位算法

Improved L1-SVD Sound Localization Algorithm in the Spherical Harmonic Domain under Reverberant Environments

由于多径效应的影响,基于稀疏重构理论的声源定位算法在混响环境下存在定位精度较差的问题.本文提出一种适用于混响环境的DPDT-L1SVD声源定位算法,该算法将L1-SVD稀疏信号重构声源定位算法扩展到球谐波域中,并根据球谐波域中导向矢量与频率解耦的特性将其拓展到宽带信号处理中,再与直接路径检测方法（direct-path dominant test,DPDT）相结合.由仿真实验结果可知,在低信噪比情况下,该算法定位精度优于DPDT-MUSIC和SS-PIV算法,且具有较强的噪声鲁棒性.结合实测实验结果,验证了该算法在混响环境下具有较高的定位精度和空间分辨率.

It is well known that the sound localization algorithm based on sparse signal reconstruction theory has poor localization accuracy under reverberant environments due to the effect of multipath distortion.In this paper,we proposed the DPDT-L1 SVD algorithm.This algorithm formulates the sound localization algorithm of L1-SVD in spherical harmonics domain,expands it to the wideband signal processing according to the characteristics of the steering vector and frequency decoupling in the spherical harmonic domain,and then takes advantage of the direct-path dominance test（DPDT）as a pre-processing step to improve the accuracy of localization under reverberation environments.The simulations demonstrate that the localization accuracy and noise robustness of the proposed algorithm are higher than those of DPDT-MUSIC and SS-PIV algorithms in low signal-to-noise ratio（SNR）.The simulations and the experiments in a real reverberant room validate the high localization precision and spatial resolution of the proposed algorithm in the presence of reverberation.